Odds That An Involved Driver Was Drinking: Best Indicator Of An Alcohol-Related Crash? A. S. Tippetts and R. B. Voas Pacific Institute for Research and Evaluation 1171 Beltsville Drive, Suite 3, Calverton, Maryland, 275 USA Keywords Impaired driving, alcohol-related crashes, driver age, driver sex, fatal crashes Abstract The evaluation of the effectiveness of new alcohol safety laws and programs is critical to the development of a successful national system for reducing alcohol-related crashes. Although evaluations of U.S. laws, such as those reducing the legal blood alcohol concentration (BAC) limit to.8 or establishing zero tolerance laws for youth 21 years old and younger, have been conducted, it often has been difficult to compare the results because of the differing outcome measures used. The measures reported include all crashes, single-vehicle crashes, nighttime crashes, had-been-drinking crashes, and crashes involving drivers with positive or high BACs. All of these measures fail to include a term reflecting the underlying rate of non-alcohol-related crashes, which can be influenced by the quality of vehicles, roadways, urbanization, and the economy, among other factors. A standard method for reporting health statistics is to use population as the normalizing measure. However, since access to a vehicle varies by socioeconomic class and urban rural location, vehicle miles of driving (VMT) appears to be the method of choice for normalizing highway injuries. However, VMT is of limited value for normalizing alcohol-related injuries because the proportion of vehicle miles of travel that involve an impaired driver is more difficult to determine and not available in the state estimates derived from state gas taxes. Conversely, nonalcohol-related crashes occurring at the same times or locations or to the same groups of drivers can provide a more specific method of normalizing alcohol-related events by controlling more specifically for vehicle type, roadway conditions, or driver characteristics. Thus, for example, the frequency of alcohol-related crash involvements to non-alcohol-related crash involvements can be compared for specific at-risk groups such as African American males ages 21 to 35, a specific group for which vehicle miles of travel are not available. This paper illustrates the use of that type of normalization in comparison vehicle miles of travel in analyzing alcohol-related crash involvements by age and gender.
Introduction A significant relationship between driver age and driver gender and involvement in impaired driving crashes has been well established (1). The reversal of the two-decade downward trend in alcohol-related fatalities in 2 (1) places renewed importance on understanding such differences for the development of new countermeasures. The best measure of alcohol s role in traffic crashes is the blood alcohol concentration (BAC) of crash-involved drivers. The Fatality Analysis Reporting System (FARS)(2) provides a national census of fatal crashes in which BACs are available for approximately 7% of killed drivers. BACs of the remaining 3% can be accurately imputed by a method developed by Klein (3) and adopted by the National Highway Traffic Safety Administration (NHTSA) for reporting FARS data. This ability to separate the drinking from non-drinking drivers in fatal crashes provides the opportunity to contrast the two for any specified group. Drinking drivers can be expressed as a percentage of all drivers for the particular group in a fatal crash or can be expressed as the odds (drinking/non-drinking) that a driver in a fatal crash will be drinking. This study compares that odds measure with the more familiar VTM measure as a method of comparing the involvement of drivers in different age and gender groups in fatal crashes. Methods From a 7-year FARS sample covering all drivers killed between January 1, 199, and December 31, 1996, a total of 162,192 drivers 117,632 (72.5%) of which had measured BACs was used in this analysis. These were divided into two groups: 66,376 with measured or imputed BACs greater than zero (drinkers) and 95,816 with zero BACs (nondrinkers). Because men and women of different ages drive varying amounts, it is necessary to normalize the crash data based on estimated miles of driving. The 1995 Nationwide Personal Transportation Survey (NPTS) conducted by the Federal Highway Administration (4) provided annual mileage estimates separately for men and women by age group. These data were used to calculate the number of drinking and nondrinking drivers in a fatal crash per billion vehicle miles traveled (VMT). A second method used the ratio of drinking drivers to nondrinking drivers within separate age/gender groups as a basis for comparing driver crash involvement. This ratio provides the odds that a driver killed in a crash had been drinking. Results The mileage death rates for male and female nondrinking drivers for different age groups are contrasted in Figure 1a. The rates for males and females were generally similar, yielding U- shaped curves with greatly elevated rates for youths under 21 and elderly drivers over 7. The greatest cross-gender difference occurred among 16- to 2-year-olds, where sober death rates were about 2 for males and about 13 for females. Figure 1b provides the mileage rates for drinking drivers. These curves contrast with those in Figure 1a in being J-curved, with the highest involvement among young drivers and with males demonstrating significantly higher rates at all age levels. The only two age/gender groups for which the mileage death rates were higher among drinking versus nondrinking drivers were males aged 21-29 and 3-39.
Figure 1a : Nondrinking driver death rates by age and gender per annual miles driven. Sober deaths (per billion VMT) 24 21 18 15 12 9 6 3 Figure 1b : Drinking driver death rates by age and gender per annual miles driven. Alcohol deaths (per billion VMT) 14 12 1 8 6 4 2 Figure 2 relates the drinking to nondrinking driver involvements by displaying the odds that a killed driver had been drinking. As expected, the odds that a driver had been drinking were greater for males than for females (odds ratios from 2.4 to 3.1, significant for all age groups with all p-values <.1) and were greatest for drivers between 21 and 39. Although differences across age/gender groups can only be inspected visually with mileage death rates (because there are no variance estimates for rates per VMT), groups can be statistically compared using odds ratios. In pairwise contrasts across age groups within each gender, the only comparisons that were not significantly different were those between the 21-29 versus 3-39 year old age groups (p-values =.21 and.78, males and females respectively; all other p-values <.1).
Figure 2. Odds that a killed driver had been drinking by age and gender. Odds 1.6 1.4 1.2 1.8.6.4.2 Discussion The U-shaped curves in Figure 1a are typical of those reported in previous studies (5). The elevated rate for underage drivers is generally attributed to inexperience and risk taking (6), while the elevated rate for the elderly is generally attributed to age-related decrements in cognitive and motor functions and increased risk of fatal injury given that a crash occurs. The most significant feature of the rate of nondrinking driver involvement is that, except for underage drivers, the rates for males and females are similar. This suggests that relative to exposure measured by VMT, females face the same risk factors as males and that traffic safety programs such as driver licensing, safety belt laws, traffic enforcement, etc. are equally effective with both genders. In contrast to Figure 1a, the drinking driver involvement rates in Figure 1b highlight the significant role that both age and gender play in alcohol-related crashes. It is important to recognize that the rates shown are based on all miles driven not just those driven at the time and places where crashes occur, as in studies using roadside surveys to calculate the relative risk of crash involvement at various BACs (7, 8). The rate shown is a function of both the sensitivity of the particular group to impairment and the number of miles driven with a positive BAC by that group, which cannot be separated from the total miles driven. Thus, the rates in Figure 1b represent the overall risk of being a drinking driver in each group based on total VMT. It is clear that the drinking driver rate is significantly lower for females and declines with age for both genders, supporting the current policy of focusing alcohol countermeasures on young males aged 21-35. Figure 2 relates the drinking to nondrinking driver rates by calculating the odds that a fatally injured driver will have been drinking. With respect to age, this gives a substantially different picture from Figure 1b. Although an underage fatally injured male driver is significantly more likely to be alcohol-free than to have been drinking, the reverse is true of an adult male driver in the 21 to 39 year age group. The relatively high rates of both sober and drinking driver deaths among 16 to 2 year olds emphasizes the need to deal with their overall risk taking and inexperience (as in the current emphasis on enacting graduated driving license laws), as well as
increasing the enforcement of underage drinking laws. The fact that the odds a fatally injured underage driver was drinking are lower than those for older drivers may, in part at least, reflect the effectiveness of current minimum drinking age (9) and zero tolerance laws (1), which have been enacted in all 5 states. Figure 2 clearly suggests that another important opportunity for further reducing alcohol-related crashes lies with countermeasures aimed at the 21 to 39 year age group, particularly among males. References 1. National Highway Traffic Safety Administration (NHTSA). Traffic safety facts 2: Alcohol. Washington, DC: National Center for Statistics & Analysis, National Highway Traffic Safety Administration; 2. 2. FARS. Fatal accident reporting system, 2. Washington, DC: National Center for Statistics and Analysis, Highway Traffic Safety Administration; 21. 3. Klein T. A method for estimating posterior BAC distributions for persons involved in fatal traffic accidents. Washington, DC: National Highway Traffic Safety Administration; 1986. DOT HS 87 94. 4. Research Triangle Institute. User's guide for the public use data files: 1995 Nationwide Personal Transportation Survey. Washington, DC: Federal Highway Administration (FHWA); 1997, October. Publication No. FHWA-PL-98-2. 5. Cirelli EC. Crash data and rates for age, sex, groups of drivers, 1994: U.S. DOT, NHTSA, National Center for Statistics and Analysis; 1995. Research Note. 6. Mayhew DR, Donelson AC, Beirness DJ, Simpson HM. Youth, alcohol and relative risk of crash involvement. Accid Anal Prev. 1986; 18:273 287. 7. Zador PL, Krawchuk SA, Voas RB. Relative risk of fatal crash involvement by BAC, age, and gender. Washington, DC: U.S. Department of Transportation, National Highway Traffic Safety Administration; 2. Technical Report No. DOT HS 89-5. 8. Borkenstein RF, Crowther RF, Shumate RP, Ziel WB, et al. The role of the drinking driver in traffic accidents. Blutalkohol. 1974; 11:1 132. 9. Toomey TL, Rosenfeld C, Wagenaar AC. The minimum legal drinking age: History, effectiveness, and ongoing debate. Alcohol Health & Research World. 1996; 2:213 218. 1. Blackman K, Voas RB, Gullberg RG, Tippetts S. Enforcement of zero tolerance in the state of Washington Evidence from breath-test records. Forensic Sci Rev. 21, July; 13:77 86.